The invention generally relates to label-free imaging systems, and relates in particular to non-linear optical microscopy and micro-spectroscopy imaging systems employing efficient dual frequency laser sources.
The development of confocal microscopy and genetically encodable fluorescent labels has transformed biological research. Labels however, may be perturbative of a sample, especially for imaging molecules that are smaller than typical fluorophores (e.g., metabolites or drugs) and that are not applicable for in vivo diagnostics due to toxicity and/or delivery concerns. Certain molecules or properties (e.g., strain or stress in material science samples) cannot be labeled at all, and photobleaching may be problematic for long term measurements.
Alternatively, vibrational spectroscopy may provide label-free chemical contrast based on intrinsic molecular properties of the sample. Yet, the traditional techniques, infrared absorption (IR) and spontaneous Raman, are limited, and IR microscopy suffers from low spatial resolution and limited imaging depth due to the long wavelength. Spontaneous Raman microscopy has slow imaging speed due to the weakness of the signal.
Micro-spectroscopy generally involves capturing a spectrum from a microscopic volume in a sample, while microscopy generally involves capturing an intensity value as well as scanning such that multiple intensity values are captured to form picture elements (pixels) of a microscopy image.
Infrared microscopy involves directly measuring the absorption of vibrationally excited states in a sample, but such infrared microscopy is generally limited by poor spatial resolution due to the long wavelength of infrared light, as well as by a low penetration depth due to a strong infrared light absorption by the water in biological samples.
Raman microscopy records the spontaneous inelastic Raman scattering upon a single (ultraviolet, visible or near infrared) continuous wave (CW) laser excitation. Raman microscopy has improved optical resolution and penetration depth as compared to infrared microscopy, but the sensitivity of Raman microscopy is rather poor because of the very low spontaneous Raman scattering efficiency (Raman scattering cross section is typically on the order of 10−30 cm2). This results in long averaging times per image, which limits the biomedical application of Raman microscopy.
Coherent Raman scattering (CRS) microscopy techniques, including coherent anti-Stokes Raman scattering (CARS) microscopy and stimulated Raman scattering (SRS) microscopy allow signal amplification by up 100,000× compared to spontaneous Raman, enabling label-free imaging with high temporal (imaging speeds up to video-rate, i.e., 30 frames/s) and sub-micron spatial resolution. Due to the use of nonlinear excitation, CRS microscopy is intrinsically three-dimensional, allowing non-destructive optical sectioning of the sample. The excitation uses near-infrared light within the optical window of biological specimen, allowing imaging depths up to a few hundred microns.
CRS is also free of photobleaching, if electronic resonances are avoided; and auto-fluorescence does not interfere, because it is not coherently amplified. CRS imaging systems may be used in biology and material science research, such as studying lipid metabolism, optimizing drug formulation for trans-dermal delivery, and in biofuel production. Label-free microscopy is also being evaluated as a medical imaging modality for delineation of tumor margins in brain and breast cancer surgery and early detection of melanoma.
Despite the advantages of CRS, high instrument cost and the technical complexity limit its wider use and currently only a few laboratories can obtain high quality images. Providing laser sources for CRS is challenging not only due to the comparative cost of laser systems as compared to a complete conventional Raman system, but the illumination must be provided as two synchronized laser pulse-trains of picosecond pulse duration, with a tunable difference frequency to the precision of a typical Raman line width.
Modulation transfer microscopy and spectroscopy imaging systems such as stimulated Raman scattering (SRS), spectral excitation of stimulated Raman scattering (SRS Spectral), stimulated emission (SE), ground state depletion (GD), photo-thermal (PT), two-color two-photon absorption (TPA), and stimulated Brillouin scattering generally involve reliance on the non-linear interaction of two laser beams within a sample, and detection of a characteristic, such as gain or loss, of one of the excitation beams. This is in contrast to detecting a newly generated (new frequency) emission signal as is done, for example, in one-photon and two-photon excited fluorescence, spontaneous Raman scattering, coherent anti-Stokes Raman scattering (CARS), second harmonic generation, (SIIG), sum frequency generation (SFG) and third harmonic generation (THG).
Such modulation transfer microscopy and micro-spectroscopy techniques require a detection scheme that provides for detection of a relatively small signal (e.g., a small gain and loss signal) on top of noisy lasers. This is generally achieved in accordance with various embodiments based on modulation transfer—by modulating a feature of one of the laser excitation beams and measuring the signal of interest with high sensitivity. In particular, the modulation transfers to the other excitation beam due to non-linear interaction within the sample, which facilitates detection of the signal of interest using a modulation sensitive detector. If the modulation frequency is chosen to be faster than the laser noise (e.g., greater than about 200 kHz), shot-noise limited sensitivity may be achieved. Such modulation schemes are readily compatible with beam-scanning microscopy and micro-endoscopy, video-rate imaging speeds, and multiplex excitation schemes.
An advantage of these non-linear optical imaging techniques as compared to fluorescence microscopy, is that they allow for specific image contrast based on intrinsic spectroscopic properties of the sample, rather than extrinsic fluorescent labeling or dye staining. This is particularly important for imaging of small molecules that can be perturbed by labeling and medical diagnostics because of a possible toxicity of the used dyes. In CARS and SRS, chemical contrast is derived from intrinsic molecular vibrations and in TPA, SE and GD microscopy from absorption properties of the molecules constituting the sample.
Common features of CARS and SRS include that each requires (1) pulsed laser beams with a pulse-width shorter than about 10 ps, i.e., a spectral bandwidth of at least about 30 GHz, and (2) two synchronized beams that are overlapped in time, i.e., the repetition rate of the lasers and the time delay between the pulse-trains must be fixed.
Modulation transfer techniques further require that a property (such as intensity, polarization or time delay) of one of the beams is modulated at a rate higher than 100 kHz allowing measurement the modulation transfer from this modulated beam to the second, originally un-modulated beam due to the nonlinear interaction in the sample
These different techniques have different laser wavelength requirements. For CARS an SRS, the difference between the two excitation frequencies (|ω1−ω2|) is selected to be resonant with a vibrational frequency of the sample. The specific wavelengths of the two excitation fields, therefore are not critical as long as the difference frequency is as desired. Such sources are typically chosen to be in the range of about 700 nm to about 1600 nm, for which biological samples are transparent. The tuning of the difference frequency to a vibrational frequency of the sample (about 200 cm−1 to about 4000 cm−1) should be to a precision of at least about 2 nm.
Stimulated emission (SE) and ground state depletion (GD) microscopy involve tuning either ω1 or ω2 to be electronically resonant with the sample. With photo-thermal (PT) microscopy, either ω1 or ω2 is chosen to match the one or two photon electronic absorption frequency. With two-color two-photon absorption (TPA), the sum of ω1 and ω2 is chosen to be electronically resonant with the sample.
Many conventional laser systems for CARS and MTM techniques have involved the use of mode-locked solid state lasers in order to achieve the pulse width shorter than 10 ps as such pulse-width that cannot conventionally be achieved with an electrically driven laser systems. A particular challenge, is the requirement of overlaps the pulses in time precisely (synchronization), as timing jitter translates into severe noise of the signal if it is bigger than the pulse width (i.e., much smaller than the required 10 ps).
Certain conventional implementations of CARS microscopy involved using two Titanium Sapphire (Ti:Sa) lasers whose outputs were electronically locked to one another using feedback regarding the cavity length of one of the lasers. Both Ti:Sa lasers were continuously tunable from about 750 nm-1000 nm, which allowed imaging based on Raman frequencies in the entire spectral region from about 200 cm−1-4000 cm−1. Such systems however, suffered from timing jitter between the pulses, making long-term experiments impossible and limiting day-to-day stability of the system.
Later developed conventional system involved the use of optical parametric oscillators (OPO) for label-free microscopy that are intrinsically locked due to synchronously pumping the OPO with the same lasers that provides the first beam. Such OPO laser systems may also be pumped with mode-locked fiber lasers. The pump laser is typically fixed at 1064 nm and the OPO output is tunable from 750 nm to 1000 nm, again allowing to image any Raman band. The long-term stability, complexity and price of such OPO laser systems however, remains a shortcoming of such systems. Moreover, dual frequency sources employing OPO laser systems typically include an adjustable translation stage that ensures that the resulting two trains of laser pulses are temporally overlapped. Such an adjustable translation stage adjusts the optical path of one of the pulse trains within a short range to ensure temporal synchronicity. Variations in temperature of the imaging system will also affect the path lengths and therefore synchronization.
Another approach to providing illumination systems for dual-excitation wavelength non-linear imaging systems has been based on time-lens lasers, which allow generation of pulses on demand with response to an electronic trigger signal. A Yb time-lens laser may be triggered by a Ti:Sa laser to provide to laser pulse trains for CRS with minimal timing jitter. One wavelength is fixed at 1040 nm and the other is tunable over the entire region of Raman spectra. Again however, the synchronization of the two oscillators is achieved electronically rather than by optical seeding.
Other approaches to providing laser pulse trains for CRS have been based on super-continuum generation (SC) in an optical fiber to generate a frequency shifted second train of pulses synchronized to the first train of pulses. Typically SC spectra are very broad, much more broad than the typical line-shape of Raman spectra and SC light sources for CRS are combined with spectral compression schemes to recover the chemical specificity of CRS. This may be achieved either by spectral focusing CRS or by spectral compression via sum frequency generation.
This permits the generation of the spectral brightness required for CRS imaging and allows for fast imaging speed with pixel dwell times as short as 4 μs. Because Raman SC spectra are very broad, SC light sources are broadly tunable (e.g., from 850 nm-1100 nm) and allow access Raman peaks across the full Raman spectrum. While relying on optical synchronization, this approach is different from the laser system disclosed herein in that the second pulse train is generated by super continuum generation in a nonlinear fiber rather than lasing; In the nonlinear fiber, the molecular population is unaffected similar to the parametric process in OPOs.
Ultrafast laser systems have also been disclosed based on seeding a Yb-doped amplified with a super-continuum generated from an Er-doped oscillator. Such systems however, only provide a single-color output, which is not suitable for CRS microscopy, which SRS requires the generation of two synchronized trains of picosecond pulses with narrow bandwidth (a few cm-1) and that are independently tunable over a wide spectral range 800-3300 cm-1. Further, they are intrinsically broadband due to the broad spectral range from the super-continuum fiber that directly seeds the amplifier.
There remains a need, therefore, for an efficient dual frequency laser system with reduced jitter for microscopy and micro-spectroscopy imaging systems.